Deep learning segmentation method based on pseudo-color CT image

Abstract

The invention discloses a deep learning segmentation method based on a pseudo-color CT image. The method comprises the following steps: step 1, acquiring a CT image data set SI0 and a corresponding label image data set SL0; 2, amplifying the data set through image rotation and zooming operation; 3, selecting three different interested window width window levels to obtain CT images under the corresponding window width window levels, and further generating a pseudo-color CT image data set SIP; 4, randomly dividing the pseudo-color CT image data set SIP and the corresponding label image data setSL0 into a training set ST and a test set SX; 5, constructing a deep learning network architecture; 6, training the deep learning network model by using the training set ST and the test set SX to obtain a trained segmentation model; and step 7, inputting image data needing to be detected into the trained deep learning network model to obtain a corresponding segmentation result. The CT image segmentation precision and efficiency can be effectively improved, and clinical diagnosis and treatment are facilitated.

Description

technical field [0001] The invention relates to the field of medical image processing, in particular to a deep learning segmentation method based on pseudo-color CT images. Background technique [0002] X-ray CT (Computed Tomography) has the advantages of fast imaging speed and high spatial resolution, and has been widely used in the fields of medical diagnosis and treatment. CT image segmentation is of great significance in 3D reconstruction, computer-aided diagnosis and treatment, radiation therapy and other fields. At present, the segmentation of the tissue of interest mostly depends on the manual outline of the doctor, but the window width and level of observing different tissue images are often quite different, and it is difficult to segment them at the same window width and level at the same time, resulting in image segmentation that not only requires a lot of time Time and energy, and easily affected by subjectivity and doctor experience, there are problems such as i...

AI Technical Summary

Problems solved by technology

However, for CT images, as well as for different tissues with large differences in observation window width and level, effective observation cannot be performed under the same window width and level, and it is difficult to achieve accurate parallel segmentation

Images